Analysing storm data using highly interactive tools

Hurricanes pose great danger to the southeastern United States, causing extensive damage to life and property. The destruction is at its worst if people are caught unprepared, so the ability to predict the path of a hurricane is a matter of great interest. In particular, it would be very useful, when a hurricane is still far out at sea, to be able to predict whether it is going to reach land, and if so, where it will strike. This paper shows an example of the use of interactive statistical graphics to study hurricane behaviour. The interactive techniques are then combined with standard parametric analysis to achieve more powerful and reliable results. The study reveals several interesting features of the dataset, e.g. hurricanes hitting countries other than the US have not been recorded as landfall by the US National Hurricane center. The approach used can be extended to study any data measured along a path, such as other kinds of storms, or tornadoes.